economic complexity and environmental performance ... · 4department of economics, university of...
TRANSCRIPT
Munich Personal RePEc Archive
Economic complexity and environmental
performance: Evidence from a world
sample
Lapatinas, Athanasios and Garas, Antonios and Boleti,
Eirini and Kyriakou, Alexandra
18 March 2019
Online at https://mpra.ub.uni-muenchen.de/92833/
MPRA Paper No. 92833, posted 23 Mar 2019 03:43 UTC
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Economic complexity and environmental per-formance: Evidence from a world sample
Athanasios Lapatinas1,∗, Antonios Garas2, Eirini Boleti3
and Alexandra Kyriakou4
1European Commission, Joint Research Centre (DG-JRC),Via E. Fermi 2749, TP 361, Ispra (VA), I-21027, Italy
2ETH Zurich, Chair of Systems Design, Weinbergstrasse 56/58, 8092 Zurich, Switzerland3TNO, Department of Climate, Air and Sustainabil-
ity, Princetonlaan 6, 3584 CB Utrecht, The Netherlands4Department of Economics, University of Ioannina, P.O. Box 1186, 45110 Ioannina, Greece
Abstract
This paper analyzes the relationship between economic complexity and environmental
performance using annual data on 88 developed and developing countries for the period of
2002-2012. We use the Economic Complexity Index, highlighting that a country’s productive
structure is associated with the amount of knowledge and know-how embodied in the goods
it produces. Measuring environmental performance through the Environmental Performance
Index, we show that moving to higher levels of economic complexity leads to better environ-
mental performance and therefore, that product sophistication does not induce environmental
degradation. Nevertheless, the effect of economic complexity on air quality is negative, i.e.,
exposure to PM2.5 and CO2 emissions increases. These findings remain robust across alter-
native econometric specifications. Furthermore, we highlight the link between the complexity
of products and environmental performance at the micro-level. We build two product-level
indexes that link a product to the average level of (a) environmental performance and (b) air
pollution (CO2 emissions) in the countries that export it. With these indexes, we illustrate
how the development of more sophisticated products is associated with changes in environ-
mental quality and show that the complexity of an economy captures information about the
country’s level of pollution.
Keywords: Economic complexity, environmental performance, pollution, CO2 emissions
∗Corresponding author ([email protected]). The final revisions of both the text and empiricalstrategy of the article were conducted when Athanasios Lapatinas took service at the European Commission’sJoint Research Centre. The scientific output expressed does not imply a European Commission policy position.Neither the European Commission nor any person acting on behalf of the Commission is responsible for anyuse that might be made of this publication.
1/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
1 Introduction
Since the industrial revolution, new technologies have radically transformed sectors/industries,
emphasizing the role of countries’ economic structures on enhancing modernization of produc-
tion. Structural transformation – the process by which economies diversify from agriculture to
more sophisticated industries and services [22, 69, 74, 78, 99] – is a process of creative destruc-
tion that directly affects the environment, by promoting for example the rapid growth of fossil
fuel consumption and by producing significant levels of carbon dioxide (CO2) emissions [80].
Although it is easy to highlight positive and negative impacts in particular sectors, it is much
more difficult to analyze and measure the overall environmental footprint of the reallocation of
activities throughout the economy.
On the one hand, the transformation of the productive structure and the process of industrial-
ization increase energy consumption and carbon emissions [82, 88, 130]. Some new production
technologies developed in recent decades have had a far greater environmental impact than the
ones they replaced. For example, in the agricultural sector, the traditional fertilizing system on
farms, where animals provided the manure to fertilize the land, has been substituted by chemical
fertilizers that have led to severe pollution problems (e.g. increase of nitrate and phosphate in
drinking water and rivers, deterioration of soil fertility) because they contain heavy metals (e.g.
cadmium and chromium) and high concentrations of radionuclides [20, 105].
On the other hand, the reallocation of factors of production from traditional to modern activities
can also have positive effects by developing new methods of reducing pollution and producing
cleaner energy (e.g., solar panels, wind turbines, hydroelectricity, etc). Green technology and eco-
innovation are decisively geared at lessening, if not reversing, the negative impacts of pollution
by creating new products or services and new management and business methods. These include
among others, innovations in renewable energy, recycling, waste-water treatment, and eco-friendly
food processing and packaging. The world market of environmental products and services is
2/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
growing and policy makers are now paying more attention to the environmental goods and
services (EGS) industry which is seen as a key ingredient of industrial competitiveness, trade
advantage and social stability [36, 65, 73, 111]
From the above discussion, it becomes clear that structural changes of the productive structure
have affected society and the environment in a number of ways. To disentangle the net effect of
the process of structural transformation on environmental performance we employ the Economic
Complexity Index (ECI), which quantifies the ‘product space’ of countries, i.e., the network rep-
resentation of the relatedness and proximity between products traded internationally. When a
country produces a good that is located in the core of the product space (such as metal products,
machinery, and chemicals) many other related goods can also be produced with the given set
of capabilities. However, this does not hold for the goods lying in the network’s periphery (e.g.,
fishing, tropical, and cereal agriculture, garments, textiles, and animal agriculture) because they
require different capabilities. The ECI methodology encapsulates this information by assigning
lower values to countries that export products located at the periphery of the product space and
higher values to countries that export commodities located in the center of the product space.
Then, the transformation of the productive structure from extractive sectors to more sophisti-
cated industries can be paralleled with the process of moving from the periphery to the core of
the international trade network [59]. In other words, the measure of economic complexity – which
we define and explain in the Appendix A – quantifies a country’s productive structure taking
into account the sophistication of the produced goods and capturing differences in industrial
structures [1, 8, 18, 19, 28, 29, 29, 40, 54, 56–59, 101, 106, 118]
In recent years, the ECI has received widespread attention throughout the scientific community,
mainly because it is a robust predictor of economic growth [55, 58]. Furthermore, Hartmann et al.
[54] have recently shown that countries exporting complex products tend to be more inclusive and
have lower levels of income inequality than countries exporting simpler products. The authors
3/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
attribute part of their finding to industrialization which played a major role in the rise of a new
middle class by creating new jobs and training/education opportunities for workers.
Industrialization has also resulted in higher levels of pollution [11, 23, 93]. The relationship
between industrialization and pollution has been investigated in numerous papers. Most of these
use cross-country data and find a positive coefficient. A typical example is the work of York et al.
[127], who show that in 146 countries, industrialization (measured as % GDP from industry)
monotonically increases energy use and CO2 emissions. Another example is Shafiei and Salim
[107], who also find a positive relationship in 29 OECD countries for the period of 1980-2011.
Asane-Otoo [13] shows that the industrialization process of middle-income African countries
had a significant positive effect on their CO2 emissions. A positive and statistically significant
relationship between industrialization and pollution is also found by Al-Mulali and Ozturk [6] in
14 Middle East and North African (MENA) countries for the period of 1996-2012. The positive
relationship between the proportion of GDP from industry and air pollution is also verified
by Martínez-Zarzoso et al. [90] for European Union member states during the period of 1975-
1999. Extensive research also exists regarding the relationship between structural change and
environmental degradation in China. Studies such as Li et al. [79], Lin et al. [84], Wang et al.
[122], Xu and Lin [126], Zhang and Lin [128] show that a country’s structural transformation
from agricultural to industrial production influences environmental quality in a negative way.
On the other hand, according to the ‘three-sector analysis’ [24, 44], as incomes rise and countries
become more prosperous, the economy moves towards the tertiary sector (service-based) and
the level of pollution decreases [32]. At this stage, demand for health, education, security and
well-being increases, as well as the level of environmental awareness. The interest on the Envi-
ronmental Kuznets Curve (EKC) has recently been revived [3, 5, 27, 31, 33, 47, 77, 91] and its
hypothesis has been used to revisit the implications of structural transformations on the nexus
between economic development and environmental performance [89].
4/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Recent advancements in information and communication technologies (ICT) have raised concerns
about ‘technological pollution’. A priori, ICT seems to be harmful for the environment because
its use implies higher energy consumption. Contrary to this perception, Romm [102] argues that
ICT supports sustainable development as the growth of ICT is linked to reductions in energy
intensity. According to Hilty et al. [60], ICT’s positive and negative impacts on environmental
sustainability take the following three forms: primary effects, such as increasing electronic waste;
secondary effects, such as improving the energy-efficiency of production; tertiary effects, such
as inducing a product-to-service shift in consumption and/or activating structural changes in
economies’ productive structures. Ollo-López and Aramendía-Muneta [95] show that the use of
some ICT helps to reduce emissions, whereas others increase them. In general, there are no
consistent results regarding the impact that ICT has on the environment, and the literature
investigating this issue is scarce.
The above-mentioned studies show a clear relationship between economic sophistication and the
environment. Here, we contribute to this literature and to recent work on economic complex-
ity by documenting a strong and robust relationship between the sophistication of a country’s
productive structure and its environmental performance. We find that higher levels of economic
complexity are associated with better environmental performance but lower air quality (i.e.,
higher exposure to PM2.5 and higher CO2 emissions). This result is consistent across various
specifications in a global sample that includes 88 developed and developing countries over the
period of 2002-2012.
Moreover, following Hartmann et al. [54], we develop two product-level indexes that allow us to
classify products according to the level of environmental performance and CO2 emissions they are
associated with. Using these indexes, we illustrate how the development of sophisticated products
is associated with environmental degradation. Our results suggest that countries’ environmental
performance is conditioned by their productive structures and hence by the level of sophistication
5/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
embodied in the products they make. Together, our indexes may be a promising policy tool that
can be used to estimate the changes in environmental performance (including air pollution) we
would expect if a country were to modify its productive structure by reshaping, for example, its
smart specialization strategy [41–43]. This policy tool could take the form of an interactive online
‘map’ in which the user can explore data/graphs of 772 products and their associated indexes
(for environmental performance and CO2 emissions) between the years 2002-2012.
The remainder of the paper is structured as follows. Section 2 discusses the datasets used in the
paper. Section 3 describes the econometric analysis for studying the effect of economic complex-
ity on environmental outcomes and discusses the control and instrumental variables included in
the econometric model. Section 4 presents the results. Section 5 introduces two indexes of the
product-level environmental performance and CO2 emissions expected for the producers and ex-
porters of 772 different products in the Standard Industrial Trade Classification at the four-digit
level (SITC-4 Rev.2). Using these indexes, we highlight the links between product sophistication
and the environment that are seen at the micro-level. We demonstrate that the development
of more complex products is associated with better environmental performance but lower air
quality (higher CO2 emissions). Finally, in Section 6, we draw our conclusions.
2 Data
Dataset 1: environmental performance We utilize the Environmental Performance Index
(EPI) which is a composite index of environmental quality developed by Yale University and
Columbia University in collaboration with the World Economic Forum and the European Com-
mission’s Joint Research Centre. The EPI includes (i) emissions indicators for different pollutants,
(ii) the effects of pollution on human health and environmental degradation, (iii) the existence
6/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
and effectiveness of environmental policies.1
The EPI ranks how well countries perform on environmental policy outcomes (larger values mean
better environmental performances) in two broad objectives including protection of human health
from environmental degradation and maintaining ecosystem vitality. National performance on
each objective is measured in nine policy categories including more than twenty proximity-to-
target indicators using weights derived from principal component analysis and expert judgment
(see Figure 1 in Hsu and Zomer [64]).
Each of the two broad and inextricably linked objectives encompasses specific environmental
policy issues. The environmental health objective includes health impacts (weight: 33%), air
quality (33%) and water and sanitation (33%). The ecosystem vitality objective includes water
resources (25%), agriculture (10%), forests (10%), fisheries (5%), biodiversity and habitat (25%),
climate and energy (25%). The nine policy categories are calculated as the weighted average
of twenty underlying proximity-to-target indicators (i.e. they measure how close countries are
to meeting internationally established targets or how they perform with respect to the best
performing countries; see Table 1 in Hsu and Zomer [64]).
We analyze the sensitivity of our baseline results using two alternative dependent variables:
(a) CO2 emissions from fuel consumption obtained from the World Bank [15], and (b) average
exposure to PM2.5, which is one of the EPI’s components. Among the different pollutants, CO2
and PM2.5 are emitted from anthropogenic sources, are directly linked to economic activities and
are considered to be two of the most serious hazards to human health at the global level [21, 34,
103]. As an alternative dependent variable, we have also used the World Bank’s sum of energy
consumption index. In this way, we additionally study the economic complexity and energy
consumption nexus, as CO2 emissions are mostly generated by the use of fossil fuels. Table 1
summarizes the definitions of the variables, the sources, and presents some descriptive statistics.
1For details on the database see http://sedac.ciesin.columbia.edu/data/collection/epi. For the methodol-ogy, see Hsu and Zomer [64].
7/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Dataset 2: exports by product We use freely available international trade data from MIT’s
Observatory of Economic Complexity (http://atlas.media.mit.edu). We choose the SITC-
4 rev.2 dataset, which provides the longest time series, combining information from a dataset
compiled by Feenstra et al. [39] for the years 1962 to 2000 and the U.N. Comtrade dataset
from 2001 to 2008 (https://comtrade.un.org), and which provides details about the products
exported by each country.
Dataset 3: economic complexity The sophistication of a country’s productive structure is
measured by its economic complexity. We measure the economic complexity of countries using
the Economic Complexity Index (ECI). The ECI quantifies the diversity and sophistication of a
country’s export structure, estimated from data connecting countries to the products they export,
and is freely available from MIT’s Observatory of Economic Complexity (http://atlas.media.
mit.ed). The index is calculated by applying the methodology described in Hausmann et al. [56]
on dataset 2.
To check the robustness of our baseline results, we also use the improved Economic Complexity
Index (ECI+). The ECI+ also measures the diversity and sophistication of a country’s export
structure, but is corrected for how difficult it is to export each product. The ECI+ is also freely
available from MIT’s Observatory of Economic Complexity. The index is calculated by applying
the methodology described in Albeaik et al. [8] on dataset 2. Albeaik et al. [8] show that the ECI+
outperforms the original ECI in its ability to predict economic growth and in the consistency of its
estimators across different econometric specifications. The ECI+ is calculated with simple linear
algebra techniques that determine the knowledge intensity of economies endogenously (from the
data) [8], recognizing that institutions, knowledge and technology are prerequisites for economic
growth. In a very recent working paper, Albeaik et al. [7] show that the ECI+ is equivalent to
the fitness complexity metric proposed by Tacchella et al. [117].
8/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
3 Regression analysis
We study the effect of economic complexity on environmental performance using datasets 1 and 3
(see Section 2). Given the availability of controls, the sample covers 88 developed and developing
countries over the period of 2002-2012.2
For the estimation, in order to control for potential endogeneity problems, we follow a fixed-effects
two-stage least squares/instrumental variables (FE 2SLS/IV) strategy.
3.1 Econometric model
We regress the baseline specification described by the following equation:
EPIi,t = α0 + β1ECIi,t + β2 log(GDP )i,t + β3[log(GDP )i,t]2 + βkcontrolsi,t + γi + δt + ui,t. (1)
Here, the environmental performance of country i in period t (EPIi,t) depends on a country’s
economic complexity (ECIi,t), which is the key regressor of our analysis, as well as the level of
economic development in per capita terms, log(GDP )i,t (taking also into account the Environ-
mental Kuznets Curve empirical findings when considering the squared term of the log(GDP )i,t
per capita [25, 53, 86, 114, 115, 120], a set of control variables described in the next subsection,
time δt and country γi fixed effects, and a stochastic term ui,t. To examine the robustness of our
results and to generalize our findings, we also replicate our analysis for (a) energy consumption,
2OECD: Australia Austria, Belgium, Canada, Chile, Czech Republic, Denmark, Estonia, Finland, France, Ger-many, Greece, Hungary, Ireland, Italy, Japan, Korea (Rep. of), Latvia Mexico, Netherlands, New Zealand, Nor-way, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey, United Kingdom, UnitedStates; non-OECD: Algeria, Azerbaijan, Bangladesh, Belarus, Bolivia, Bosnia and Herzegovina, Brazil, Bulgaria,Cameroon, China, Colombia, Costa Rica, Croatia, Dominican Republic, Ecuador, Egypt, Arab Rep., El Sal-vador, Georgia, Ghana, Guatemala, Honduras, India, Indonesia, Iran, Jamaica, Jordan, Kazakhstan, Kenya,Kuwait, Lebanon, Lithuania, Macedonia FYR, Malaysia, Moldova, Morocco, Nigeria, Oman, Pakistan, Panama,Paraguay, Peru, Philippines, Qatar, Romania, Russian Federation, Saudi Arabia, Serbia, Singapore, South Africa,Sri Lanka, Thailand, Tunisia, Ukraine, Uruguay, Uzbekistan, Venezuela.
9/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
(b) air pollution and (c) CO2 emissions. In addition, we substitute the ECI with the ECI+,
finding qualitatively similar results. In Section 4, we present and discuss our findings.
3.2 Control Variables
Based on previous literature, we include in the estimated equation a number of control variables
that are likely related to environmental performance. First, in order to account for different
stages of economic development, we use the logarithm of GDP per capita (PPP in constant 2011
international dollars) from the World Bank’s World Development Indicators (WDI). According
to the EKC hypothesis, there is an inverted U-shaped relationship between pollution indicators
and economic development. To control for this, we add as an explanatory variable the quadratic
specification of GDP per capita, GDP per capita2. Since the variable GDP per capita is highly
correlated with its squared term, it has been demeaned to have a zero sample mean in order to
reduce the collinearity problem (i.e., we center the linear term around its sample mean before
taking the square) [4, 86, 100].
In addition, a nation’s openness to international trade may have a significant impact on envi-
ronmental quality. Studies on the relationship between trade and the environment give evidence
of ambiguous effects [12, 26, 38, 45, 51, 66, 67, 71]. In order to investigate whether openness
to trade influences environmental performance, we use the proportion of exports and imports
in GDP, denoted as trade. Moreover, we control for the proportions of both agriculture and
industry’s value added in GDP to capture the composition of a country’s output, denoted as
agriculture and industry, respectively. We also include two demographic variables: population
density (number of people per square kilometer of land area) and urban population (the pro-
portion of urban population) to identify the consequences of demographic changes on the en-
vironment [61, 68, 81, 97, 110, 121]. Poor air and water quality, insufficient water availability,
waste-disposal problems, and high energy consumption are exacerbated by the increasing pop-
10/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Table 1: Variable definitions, sources and summary statisticsVariable Definition Source Mean Std. Dev.
EPI Environmental Performance Index Yale Center for Environmental Law and Policy (YCELP) 49.48 16.52
ECI Economic Complexity Index: measure of the di-versity and sophistication of a country’s exportstructure
MIT’s Observatory of Economic Complexity -0.004 0.999
ECI+ Improved measure of Economic Complexity Index.See Albeaik et al. [8]
MIT’s Observatory of Economic Complexity -0.008 0.995
GDP percapita
(log) GDP per capita, PPP (constant 2011 inter-national $)
World Development Indicators 1.589 1.654
education Enrollment in secondary education, both sexes(number)
World Development Indicators 3.3M 11.6M
trade Imports plus exports as % of GDP World Development Indicators 89.93 48.25
agriculture Agriculture, value added (% of GDP) World Development Indicators 13.92 12.88
industry Industry, value added (% of GDP) World Development Indicators 29.52 12.57
population Population density (people per sq. km of landarea)
World Development Indicators 159.73 528.97
corruption Re-scaled control of corruption index. Higherscores correspond to better institutions.
Worldwide Governance Indicators 2.41 1.007
patents (log) Number of patents granted as distributed byyear of patent grant
U.S. Patent and Trademark Office 3.37 2.85
articles (log) Number of scientific and technical journalarticles
National Science Foundation, Science and Engineering Indicators 5.64 2.97
energy (log) Energy use (kg of oil equivalent per capita) World Development Indicators 7.19 1.15
CO2 (log) CO2 emissions (kg per 2011 PPP $ of GDP) World Development Indicators -1.53 0.69
air quality EPI - Issue Category: Air Quality (Weighting0.33)
Yale Center for Environmental Law and Policy (YCELP) 79.21 16.29
urban Urban population (% of total) World Development Indicators 54.91 22.97
popgrow Population growth (annual %) World Development Indicators 1.55 1.65
popold Population aged 65 and above (% of total) World Development Indicators 7.39 5.13
politicalcorruption
Political corruption index. Higher values reflecthigher levels of corruption.
Varieties of Democracy Dataset version 6.2 0.524 0.279
rural Rural population (% of total population) World Development Indicators 44.74 23.01
tertiaryeducation
Gross enrollment ratio, tertiary, both sexes (%) World Development Indicators 35.31 26.67
economicglobalization
Actual flows (trade, foreign direct investment,stocks, portfolio investment, income paymentsto foreign nationals), restrictions (hidden importbarriers, mean tariff rate, taxes on internationaltrade, capital account restrictions). Higher valuesreflect greater economic globalization.
KOF Index of Globalization 60.35 17.04
politicalglobalization
Embassies in country, membership in internationalorganizations, participation in UN security coun-cil missions, international treaties. Higher valuesreflect greater political globalization.
KOF Index of Globalization 64.15 20.94
quality ofgovernment
ICRG Indicator of Quality of Government. Highervalues indicate a higher quality of government.
International Country Risk Guide - The PRS Group 0.526 0.204
shadow Level of the shadow economy Elgin et al. [35] 32.39 12.68
executiveconstraints
Executive constraints (decision rules) Polity IV Annual Time-Series, 1800-2015 5.04 2.02
11/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
ulation density and demands of urban areas. In many countries, most cities are growing at a
faster rate than the national average, which puts pressure on urban resources and the envi-
ronment. In developing countries, workers are migrating from rural to urban areas for better
services and this could be considered an additional source of pollution [71]. On the other hand,
urbanization may increase environmental awareness and lead to a more efficient exploitation of
energy and natural resources. It is therefore possible for more densely populated areas to be less
polluted [63, 109, 119].
We also control for the level of human capital by employing the number of enrolled people in sec-
ondary education. This variable is expected to affect environmental sustainability because people
with more education are more aware of and concerned about environmental hazards. They also
tend to engage in actions that promote and support political decisions that protect the environ-
ment [16, 17, 98]. Finally, we include data on the control of corruption. The relevant literature
distinguishes two partial effects that corruption may have on environmental pollution. On the
one hand, corruption directly increases pollution by reducing the stringency of environmental
regulations [30, 87] and/or the effectiveness with which they are enforced [50, 85].3 On the other
hand, corruption affects pollution indirectly by reducing output [52, 72, 92], which in turn may
lead to lower pollution at some income levels [49, 72, 108]. This implies that at the aggregate
level, the size of the indirect effect could dominate the direct effect, leading to overall pollution
going down [25, 48]. In other words, the total effect of corruption on environmental quality can-
not be determined a priori [124]. We adopt the measure of control of corruption provided by the
World Bank’s Worldwide Governance Indicators (WGI). The index has been re-scaled to range
from 0 to 5.0, with higher scores corresponding to better institutions.
In subsection 4.2, we verify the robustness of our results by adopting further control variables
and/or using alternative measures for some of the above-mentioned controls.
3See also Lapatinas et al. [76] who explore additional channels through which corruption may impact environmentalquality in a theoretical model. See also the discussion in Lisciandra and Migliardo [86].
12/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
3.3 Instrumental variables
We estimate equation (1) using different econometric methods. First, we use pooled-OLS and
then, fixed-effects-OLS. However, fixed effects estimators do not necessarily identify the effect
of economic complexity on environmental performance. The estimation of the effect requires
exogenous sources of variation. While we do not have an ideal source of exogenous variation
recognized by previous studies, there are two promising potential instruments of ECI that we
adopt in our fixed-effects 2SLS/IV analysis.
Firstly, we use the measure of the (log) number of journal articles published in scientific and
technical journals in a given year. This index calculates the total number of papers in the fields
of physics, biology, chemistry, mathematics, clinical medicine, biomedical research, engineering
and technology, and earth and space sciences. Higher values are associated with a higher level
of scientific effort and output, which is directly related to the intensity of process and product
innovation in the economy i.e. to the sophistication of its productive structure.
The second instrument considered is the (log) number of patents granted per year from the
US Patent and Trademark Office. Both variables are expected to be correlated with economic
complexity without having a direct effect on environmental performance and this is validated by
our estimations.
We expect articles and patents to increase technological capabilities which, in turn, influence
products’ sophistication and industrialization (see Lall [75]). Regarding excludability of the in-
struments, while we do not have a precise theory for why articles and patents should have no
direct effect on a country’s environmental footprint, it is expected that changes in the number
of journal articles and patents impact environmental performance only indirectly, through the
enhanced sophistication of production.
13/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
4 Empirical results
4.1 The effect of economic complexity on environmental performance
In this section, we discuss our baseline findings, i.e., the results when estimating equation (1)
with different econometric techniques. Table 2 reports the results of pooled-OLS, adding an ad-
ditional variable from the set of controls in each step (column). In all specifications we consider
time fixed-effects. In all columns except (1), we adopt a set of regional dummies for geographical
heterogeneity, which is related to latitude, climatic conditions, and ecological awareness. Namely,
we use the following dummies: europe, asia, oceania, north america, south america. In column
(6), we also adopt the dummy variable OECD to isolate the effect of high levels of economic
development on environmental quality [46, 83, 86]. As expected, the sign of the estimated coef-
ficient is positive because of the higher environmental awareness in developed countries. In all
specifications, economic complexity has a positive relationship with environmental performance
and the control variables enter with the expected sign. The education coefficient is negative,
though its magnitude is negligible.
In columns (1)-(5) of Table 3, we estimate equation (1) with fixed-effects OLS panel regressions.
We use time dummies and robust standard errors (in parentheses). In all cases, the ECI is a
positive and statistically significant predictor of environmental performance. The statistically
significant and positive squared term of the (log) GDP per capita predicts that environmental
quality improves at higher incomes, which is in accordance with the EKC hypothesis that the
relationship between pollution and economic development follows an inverse U-shaped form [49,
62, 108]. Notably, when we control for economic complexity, the declining part of the EKC is even
more pronounced than when we do not (column 6). In addition, the role of industry in terms of
value added as % of GDP seems to be statistically important for environmental quality. Together,
all variables explain 57.6% of the variance in environmental performance among countries and
14/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
across time (column 4), but ECI is the most significant variable in the regression analysis, and
it is also the variable that explains the largest percentage of the variance in environmental
performance after the effects of all other variables have been taken into account. The semi-
partial correlation of ECI (the difference in R-squared between the full model and one in which
only ECI was removed) is 8.2%, meaning that 8.2% of the variance in environmental performance
– which is not accounted for by the other macroeconomic variables – is explained by ECI. This
in turn implies that ECI contains information about environmental performance that cannot be
explained by these other variables.
The fixed-effects 2SLS/IV results in column (6) verify that the effect of a country’s economic
complexity on its environmental performance is positive. In fact, we find that an increase in
economic complexity of one standard deviation is associated with an improvement of 5.5 in the
EPI (standard deviation: 16.5). This positive effect of economic complexity on environmental
performance is robust to the inclusion of measures of GDP per capita, population density (people
per square kilometer of land area), agriculture value added (% of GDP), industry value added
(% of GDP), control of corruption, trade (% of GDP), urban population (% of total), and
enrollment in secondary education. In the fixed-effects 2SLS/IV estimation we report (a) the
F -test for the joint significance of the instruments in the first stage: the rule of thumb is to
exceed 10 [116]; (b) the Durbin-Wu-Hausman (DWH) test of endogeneity of regressors: the null
hypothesis that the IV regression is not required is rejected; (c) the Cragg-Donald F-statistic
(Weak-id) that tests the relevance of the instruments in the first-stage regression: no evidence
of instruments having a low correlation with the endogenous regressor after controlling for the
exogenous regressors; (d) the Kleibergen-Paap Wald test (LM -weakid) of weak identification:
the null hypothesis that the model is weakly identified is rejected; (e) the p-value of Hansen’s
test of overidentification: the acceptance of the null indicates that the overidentifying restrictions
cannot be rejected.
15/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Concluding this section, the above analysis suggests that countries with more sophisticated pro-
ductive structures that lie in the core of the international trade network of products, tend to
have significantly higher environmental performance than countries in the periphery of the net-
work exporting simple products. Furthermore, exploiting the temporal variation in the data, the
fixed-effects panel regression and the fixed-effects 2SLS/IV analysis reveal a positive, statistically
significant and robust effect of economic complexity on environmental performance.
Table 2: The effect of economic complexity on environmental performance: pooled OLS(1) (2) (3) (4) (5) (6) (7) (8)
ECI 6.414*** 4.869*** 5.167*** 3.904*** 3.584*** 3.459*** 4.071*** 3.220***(0.331) (0.337) (0.343) (0.378) (0.408) (0.403) (0.422) (0.39)
GDP per capita 7.805*** 7.770*** 7.532*** 7.704*** 6.891*** 7.158*** 6.541*** 5.760***(0.321) (0.306) (0.305) (0.437) (0.478) (0.476) (0.584) (0.576)
GDP per capita2 0.443*** 0.639*** 0.658*** 0.883*** 0.651*** 0.591*** 0.14 -0.262(0.136) (0.149) (0.149) (0.157) (0.169) (0.165) (0.182) (0.179)
population -0.006*** -0.006*** -0.006*** -0.006*** -0.005*** -0.006***(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
agriculture -0.115*** -0.106*** -0.140*** -0.129*** -0.157***(0.038) (0.038) (0.039) (0.046) (0.044)
industry -0.057*** -0.028 -0.049** 0.005 0.012(0.022) (0.022) (0.023) (0.027) (0.026)
corruption 1.248*** 1.036*** 2.144*** 1.186***(0.377) (0.371) (0.383) (0.354)
trade 0.023*** 0.006 0.015***(0.005) (0.005) (0.005)
urban 0.028 0.026(0.02) (0.02)
education -0.000*** -0.000***(0.000) (0.000)
OECD 6.523***(0.774)
Observations 1283 1210 1210 1160 1160 1149 940 940R-squared 0.814 0.855 0.857 0.865 0.866 0.87 0.89 0.9F-statistic 555.8 525.3 521.8 479 466.7 460.2 483.9 526.2Note: Dependent variable: Environmental Performance Index (EPI). Main independent variable: Economic Complex-ity Index (ECI). Time fixed effects are included in all regressions. Regional dummies are also included: europe, asia,oceania, north america, south america. Robust standard errors in parentheses. * p<0.10, ** p<0.05, *** p<0.01
16/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Table 3: The effect of economic complexity on environmental performance: baseline results
(1) (2) (3) (4) (5) (6)FE OLS FE OLS FE OLS FE OLS FE OLS FE 2SLS/IV
ECI 0.650*** 0.765*** 0.778*** 1.042*** 5.519***(0.233) (0.257) (0.256) (0.317) (1.46)
GDP per capita 0.177 0.32 0.326 0.121 0.752 -3.069***(0.772) (0.572) (0.595) (0.762) (0.713) (1.023)
GDP per capita2 0.498* 0.674*** 0.782*** 0.563* 0.081 0.770**(0.297) (0.239) (0.267) (0.311) (0.38) (0.382)
population -0.002*** -0.002*** -0.002*** -0.008 -0.007 -0.006**(0.001) (0.001) (0.001) (0.006) (0.005) (0.003)
agriculture 0.016 0.007 0.016 0.016 -0.116**(0.021) (0.023) (0.033) (0.023) (0.053)
industry 0.013 0.011 0.030* 0.043** 0.023(0.015) (0.015) (0.018) (0.02) (0.025)
corruption -0.14 -0.025 -0.387 0.359(0.33) (0.347) (0.414) (0.352)
trade 0.005 0.002 0 -0.009(0.005) (0.005) (0.003) (0.007)
urban -0.088 -0.037 -0.223***(0.061) (0.056) (0.055)
education -0.000 -0.000 -0.000(0.000) (0.000) (0.000)
Fist-stage resultspatents 0.050***
(0.013)articles 0.128***
(0.036)
Observations 1283 1227 1216 985 1394 736Countries 117 114 114 110 160 88R-squared 0.495 0.552 0.557 0.576 0.494 0.417F-test 13.09DWH-test 13.94Weak-id 20.88LM-weakid 25.08Hansen (p-value) 0.8Note: Dependent variable: Environmental Performance Index (EPI). Main independent variable: EconomicComplexity Index (ECI). All regressions include time dummies. Robust standard errors in parentheses. F-testgives the F-statistic for the joint significance of the instruments in the first stage. DWH-test is the Durbin-Wu-Hausman test of endogeneity of the regressors. LM-weakid gives the Kleibergen-Paap Wald test of weakidentification. Weak-id gives the Cragg-Donald F-statistic for weak identification. Hansen (p-value) gives thep-value of the Hansen test of overidentification. * p<0.10, ** p<0.05, *** p<0.01
17/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
4.2 The effect of economic complexity on environmental performance: sensi-
tivity analysis
In this subsection, we investigate the robustness of our baseline findings. First, we use an alterna-
tive measure of economic complexity, namely the improved Economic Complexity Index (ECI+)
developed by MIT’s Observatory of Economic Complexity. The ECI+ outperforms the original
ECI in its ability to capture the difficulty of exporting each product. Second, we investigate
whether the effect of economic complexity on environmental performance survives under addi-
tional and/or alternative control measures. Third, we consider the effect of economic complexity
on energy consumption, which is a substantially interesting issue. As discussed in Section 1,
an increasing number of empirical studies investigate the relationship between industrialization
and energy use, typically finding a positive effect. Fourth, considering the EPI’s component of
air quality as the dependent variable, which accounts equally for household air quality, average
exposure to PM2.5 and PM2.5 exceedance, we replicate our baseline analysis studying the effect
of economic complexity on this measure of air pollution. However, since CO2 emissions is the
most widely used measure of pollutant emissions in the literature, we also estimate the main
specification using this measure as the dependent variable, backing up our previous results.
18/46
A.Lapatin
as,
A.G
ara
s,E
.B
oleti
and
A.K
yria
kou:
econom
icco
mplex
ityand
enviro
nm
enta
lperfo
rmance
Table 4: The effect of economic complexity on environmental performance: robustness checks(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
ECI 5.507*** 5.997*** 5.883*** 6.026*** 5.960*** 5.903*** 6.497*** 2.603** 6.584*** 3.715**(1.475) (1.531) (1.436) (1.551) (1.589) (1.706) (1.702) (1.099) (1.818) (1.893)
ECI+ 9.420***(2.362)
GDP per capita -4.082*** -3.169*** -3.149*** -2.565*** -3.198*** -2.935*** -2.911*** -3.625*** -0.499 -3.422*** -1.677(1.174) (1.042) (1.085) (0.988) (1.051) (0.988) (0.995) (1.215) (1.04) (1.202) (1.198)
GDP per capita2 1.529*** 0.948** 0.843** 0.657 0.778* 0.67 0.67 0.885** 1.452*** 0.761* 1.114*(0.464) (0.387) (0.408) (0.428) (0.411) (0.421) (0.426) (0.439) (0.478) (0.434) (0.452)
popgrow 0.108* 0.077(0.064) (0.205)
popold 0.063(0.122)
political corruption -0.045 -1.137(1.146) (1.471)
rural 0.231*** 0.098*(0.059) (0.053)
economic globalization -0.011 -0.011 0.009(0.018) (0.017) (0.027)
political globalization -0.002 -0.000(0.018) (0.019)
quality of government -3.378** -0.762(1.69) (1.410)
shadow -0.173** -0.244**(0.073) (0.113)
executive constraints -0.319 0.136(0.23) (0.177)
tertiary education 0.017(0.015)
Observations 734 734 693 682 693 691 691 678 444 685 403Countries 88 88 84 81 84 83 83 80 74 83 63R-sq 0.427 0.413 0.377 0.422 0.380 0.380 0.384 0.349 0.600 0.355 0.604F-test 13.60 13.07 13.09 14.95 12.62 12.14 12.05 13.05 13.65 11.48 7.9DWH-test 14.33 14.61 16.43 20.55 16.03 15.85 13.70 17.32 1.285 16.14 2.504Weak-id 21.69 20.93 20.49 22.44 19.94 18.97 17.42 19.17 21.52 16.16 10.71LM-weakid 22.72 24.88 24.61 27.76 24.26 24.14 24.51 24.82 21.41 23.51 13.05Hansen (p-value) 0.524 0.884 0.935 0.438 0.910 0.749 0.705 0.643 0.514 0.661 0.600Note: Dependent variable: Environmental Performance Index (EPI). Main independent variable: Economic Complexity Index (ECI). ECI+ is the improvedmeasure of economic complexity (see text). Regression analysis: FE 2SLS/IV, ECI+ is instrumented. To save space, the first stage results are not includedin the table. These are available upon request. All regressions include time dummies and the set of controls used in the benchmark specification (Table 3).Robust standard errors in parentheses. F-test gives the F-statistic for the joint significance of the instruments in the first stage. DWH-test is the Durbin-Wu-Hausman test of endogeneity of the regressors. LM-weakid gives the Kleibergen-Paap Wald test of weak identification. Weak-id gives the Cragg-DonaldF-statistic for weak identification. Hansen (p-value) gives the p-value of the Hansen test of overidentification. * p<0.10, ** p<0.05, *** p<0.01
19/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Table 5: The effect of economic complexity on environmental performance: alternative environ-mental measures
(1) (2) (3)energy air-quality CO2
ECI 0.147* -6.209* 0.311**(0.085) (3.531) (0.133)
GDP per capita 0.644*** -3.122 -0.139(0.072) (2.672) (0.109)
GDP per capita2 -0.084*** 3.344*** -0.161***(0.024) (1.18) (0.04)
population 0.000 -0.028* 0.001*(0.000) (0.015) (0.000)
agriculture 0.012*** -0.336*** 0.016***(0.003) (0.125) (0.006)
industry 0.008*** -0.014 0.008***(0.002) (0.065) (0.002)
corruption 0.022 -3.622*** 0.068**(0.02) (0.863) (0.035)
trade 0.000 0.037** -0.001(0.000) (0.016) (0.001)
urban 0.008** -0.166 0.006(0.003) (0.104) (0.004)
education 0.000 -0.000*** 0.000(0.000) (0.000) (0.000)
Observations 736 736 736Countries 88 88 88R-squared 0.465 0.366 0.364F-test 13.09 13.09 13.09DWH-test 4.697 4.745 7.94Weak-id 20.88 20.88 20.88LM-weakid 25.08 25.08 25.08Hansen (p-value) 0.332 0.135 0.143Note: Dependent variable: as noted in columns. Main independent variable: Economic ComplexityIndex (ECI). Regression analysis: FE 2SLS/IV. To save space, the first stage results are notincluded in the table. These are available upon request. All regressions include time dummies.Robust standard errors in parentheses. F-test gives the F-statistic for the joint significance ofthe instruments in the first stage. DWH-test is the Durbin-Wu-Hausman test of endogeneity ofthe regressors. LM-weakid gives the Kleibergen-Paap Wald test of weak identification. Weak-idgives the Cragg-Donald F-statistic for weak identification. Hansen (p-value) gives the p-value ofthe Hansen test of overidentification. * p<0.10, ** p<0.05, *** p<0.01
20/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Improved measure of economic complexity (ECI+)
Column (1) in Table 4 reports the estimates using the baseline fixed-effects 2SLS/IV specification
and ECI+ as an alternative measure of economic complexity (the regression includes the set of
controls used in the benchmark specification and time dummies). The baseline results remain
qualitatively intact. Particularly, the coefficient of ECI+ is positive and statistically significant
in the instrumented regression. On average, keeping all other variables constant at their mean
values, an increase of 1 point in the ECI+ increases the EPI by 9.4 points. The level of devel-
opment, measured by (log) GDP per capita, again shows a non-linear impact on environmental
performance. The EKC hypothesis also appears to be affirmed with the ECI+ as explanatory
variable. The negative coefficient of the GDP per capita variable combined with the positive sign
of its squared term confirms the inverse U-shaped relationship between pollution and economic
development.
Alternative/additional controls
Columns (2)-(10) in Table 4 start from the benchmark specification with the full set of controls
(column (6) of Table 3) and introduce additional variables or alternative measures for some of
the previous controls. Specifically, in column (2) we substitute the population density variable
with population growth (%), popgrow. Another alternative measure of population density that
also captures differences in countries’ demographic characteristics is employed in column (3),
namely the proportion of the total population aged 65 and above, popold. In column (4) the
corruption variable was substituted by the political corruption index found in the ‘Varieties of
Democracy Dataset’, version 6.2. In column (5), we use rural population (% of total) instead of
urban population, finding the opposite sign (a positive effect), as expected. Columns (6) and (7)
employ two measures of globalization, namely economic globalization and political globalization
(from the KOF index of globalization) to control for countries’ openness instead of trade. Column
21/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
(8) introduces the International Country Risk Guide (ICRG) index of quality of government,
establishing the robustness of our findings to the use of institutional quality measures. The
coefficient has a negative sign, but when the control variables are considered all together (column
(11)), the statistical significance disappears. In column (9), we adopt a measure of the level of
shadow economy, which seems to have a negative relationship with environmental performance.
In column (10), the executive constraints variable, from the Polity IV Project, is an alternative
measure of institutional quality. Finally, in column (11) we estimate the baseline model (a)
considering all of the control variables together and (b) substituting the number of people enrolled
in secondary education with the gross enrollment ratio (%) in tertiary education. Adding these
controls in our estimations leaves the findings qualitatively and quantitatively intact.
Energy consumption
The relationship between economic growth, environmental pollution and energy consumption has
been studied thoroughly in recent decades, using data from different countries and regions. Most
studies are for single countries [9, 10, 51, 112, 113, 123, 129] and only a few papers have used multi-
country data to investigate this relationship, producing ambiguous results [2, 38, 71, 96, 104].
Column (1) of Table 5 reports the estimation results of the benchmark specification but using
(log) energy consumption (kg of oil equivalent per capita) as the dependent variable. Economic
complexity has a positive effect on energy use and the same effect stands for the level of income.
However, for higher stages of economic development, it seems that energy consumption is less
intensive (the squared term of GDP per capita has a negative coefficient). Both agriculture and
industry sectors seem to be energy demanding and as expected, a higher proportion of urban
population is associated with higher energy consumption.
22/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Air pollution
As discussed in Section 2, the EPI is a composite index that ranks countries’ performance in the
following two broad policy areas: (a) protection of human health from environmental harm and
(b) vitality of ecosystems. The first component accounts for 40% of the EPI’s total score and
includes health impacts (33%), air quality (33%) and water and sanitation (33%). In this subsec-
tion, we check the robustness of our results by re-running the benchmark fixed-effects 2SLS/IV
regression with the air quality component of EPI as the dependent variable. In this way, we
focus on the effect of economic sophistication on air quality, leaving out of the equation the hu-
man well-being factors and the other “non-emission” variables. The environmental “non-emission”
components are more liable to benefits from increased income, therefore the positive effect of eco-
nomic complexity on environmental performance might derive from these components, as well as
their relatively higher weight in the comprehensive EPI. Column (2) of Table 5 shows that the ef-
fect of economic complexity on air quality is negative (at the 10% level of statistical significance).
The EKC hypothesis is verified again by the statistically significant positive coefficient of the
squared GDP per capita. Regarding the rest of the control variables, population, agriculture
and corruption enter the equation with a negative sign, while trade seems to have a positive
effect.
CO2 emissions
Column (3) of Table 5 compares the finding above with the result when the logarithm of CO2
emissions (the most commonly used indicator of air pollution in the literature) is used. The effect
of economic complexity on CO2 emissions is positive and statistically significant at the 5% level.
The squared term of GDP per capita is negative and population, agriculture, industry and
corruption all have a statistically significant positive sign. Therefore, it is verified that although
economic complexity has a positive effect on the composite and comprehensive measure of a
23/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
country’s environmental performance, this is not the case for air quality. Explicitly, it seems that
when an economy accelerates from an agricultural productive structure to a more sophisticated
one with industrial and technological sectors, the effect on overall environmental performance is
positive but the particular effect on air quality is negative.
5 Product sophistication and the environment
Using the ECI methodology, Hartmann et al. [54] recently introduced a measure that associates
products with income inequality and showed how the development of sophisticated products is
associated with changes in income inequality. Here, we introduce a measure that links a product
to the average environmental performance and air pollution level of the countries that export
it. In this way, we illustrate how environmental performance is being affected by the level of
sophistication of exported products and we quantify the influence of countries’ level of economic
complexity on their environmental performance.
Following the methodology in Hartmann et al. [54], we define the Product Environmental Per-
formance Index (PEPI) and the Product Air Pollution Index (PAPI) as the average EPI and
the average level of CO2 emissions, respectively, of the countries that export the focal product,
normalized by the importance of this product to the total exports of the countries that export
it. More precisely, we decompose the relationship between the ECI and both the EPI and CO2
emissions into individual economic sectors by creating product-level estimators of these measures
for the countries exporting a given product.
5.1 Product environmental indexes
Assuming that we have trade data for l countries and k products, we can fill the (l × k) matrix
M so that its matrix element Mcp = 1 if country c has a RCA for product p, and zero otherwise
24/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
−3 −2 −1 0 1
30
40
50
60
70
<PCI>
<P
EP
I>
−3 −2 −1 0 1
0.1
0.2
0.3
0.4
0.5
0.6
<PCI>
<PA
PI>
Figure 1: PEPI and PAPI against PCI. The solid lines represent the fit of a linear model and thedashed lines a 95% prediction interval based on the fitted linear model.
(see the Appendix A). Dataset 2 (see Section 2) contains information for the 88 developed and
developing countries used in the above analysis and for 772 products from 2002 to 2012, classified
according to the Standard International Trade Classification (SITC) at the 4-digit level.
Every product p generates some value for the country c that exports it. Therefore, for every
product p, we can calculate the fraction scp:
scp =Xcp∑p′ Xcp′
, (2)
where Xcp is the total export value of product p when exported by country c, while∑
p′ Xcp′ is
the value of all exports of country c. If EPIc (resp. CO2c) is the EPI (resp. CO2 emissions) of
country c, we can calculate the PEPIp and the PAPIp for every product as:
PEPIp =1
Np
∑
c
McpscpEPIc, (3)
PAPIp =1
Np
∑
c
McpscpCO2c, (4)
where Np =∑
cMcpscp is a normalization factor.
25/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Table 6: List of the five products with the highest and lowest PEPI and PAPI values duringthe period of 2002-2012
SITC4 Product name Product section PEPI PAPI PCI
Highest PEPI
8851 Watches and clocks Miscellaneous manufactured articles 77.2 0.16 0.455415 Medicinal and pharmaceutical products Chemicals and related products, n.e.s. 76.9 0.19 1.207259 Paper mill and pulp mill machinery, paper-cutting machines Machinery and transport equipment 76.2 0.26 1.235416 Glycosides, glands, antisera, vaccines Chemicals and related products, n.e.s. 75.9 0.21 1.187913 Railway vehicles Machinery and transport equipment 75.3 0.23 1.01
Lowest PEPI
2654 Vegetable textile fibres, waste of these fibres Crude materials, inedible, except fuels 32.0 0.08 -0.962923 Vegetable materials of a kind used primarily for plaiting Crude materials, inedible, except fuels 30.9 0.12 -0.782631 Cotton (other than linters), not carded or combed Crude materials, inedible, except fuels 30.8 0.18 -2.822683 Fine animal hair, not carded or combed Crude materials, inedible, except fuels 29.4 0.16 -0.472922 Lac, natural gums, resins, gum resins, and balsams Crude materials, inedible, except fuels 26.0 0.11 -2.01
Highest PAPI
6812 Platinum metals Manufactured goods classified chiefly by material 54.0 0.60 0.737915 Railway vehicles Machinery and transport equipment 56.0 0.60 0.493353 Pitch and pitch coke Mineral fuels, lubricants 59.1 0.57 0.362814 Roasted iron pyrites Crude materials, inedible, except fuels 55.2 0.54 0.046712 Pig-iron and spiegeleisen Manufactured goods classified chiefly by material 49.7 0.53 0.07
Lowest PAPI
2922 Lac, natural gums, resins, gum resins, and balsams Crude materials, inedible, except fuels 26.0 0.11 -2.012225 Sesame seeds Crude materials, inedible, except fuels 34.0 0.10 -2.584314 Vegetable waxes Animal and vegetable oils, fats and waxes 39.7 0.10 -1.292876 Tin ores and concentrates Crude materials, inedible, except fuels 36.5 0.09 -1.442654 Vegetable textile fibres, waste of these fibres Crude materials, inedible, except fuels 32.0 0.08 -0.96
Notes: PEPI: Product Environmental Performance Index; PAPI: Product Air-Pollution Index. Average values for2002-2012
For every year in the period of 2002-2012, we utilize the EPI and CO2 emissions for the 88
developed and developing countries in our sample and calculate the mean value of all the product-
related indexes (PCI, PEPI and PAPI) for each product.
Table 6 lists the five products with the highest and lowest PEPI and PAPI values during the
period of 2002-2012. It is evident that technologically moderate manufacturing industries and
primary sectors such as textile fibres and their wastes (SITC Rev.4 division: 22) and crude
animal and vegetable materials (SITC Rev.4 division: 29) appear to be associated with lower
environmental performance. Regarding the index of air pollution (PAPI), the reader can easily
verify that a country’s RCA in products such as non-ferrous metals (SITC Rev.4 division: 68)
and railway vehicles (SITC Rev.4 group: 791) is associated with higher CO2 emissions. In Table
6, the SITC Rev.4 group 2654 of vegetable textile fibres and the waste of these fibres and the
26/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
SITC Rev.4 subgroup 2922 of lac, natural gums, resins, gum resins, and balsams appear to be
environmentally detrimental but air pollution friendly.
5.2 Product environmental indexes and product complexity
We test the existence of a bivariate relationship between the PCI and the PEPI and PAPI
indexes by calculating Pearson’s correlation coefficient for both pairs, PEPI against PCI and
PAPI against PCI. If such a relation exists, it should allow us to derive expectations about
whether product sophistication can be associated with the quality of the environment and the
level of CO2 emissions. For the case of PEPI against PCI, the correlation coefficient is ρ = 0.74
with a p-value < 10−99, while for PAPI against PCI, it is ρ = 0.29 with a p-value < 10−15. In
Figure 1, we present the scatter plots of the PEPI and PAPI indexes against the PCI index for
all 772 products in our dataset, together with the fitted linear models. The slopes of the linear
fits are the corresponding correlation coefficients.
The statistically significant positive correlations between (a) PEPI and PCI and (b) PAPI and
PCI, indicate that more sophisticated products are associated with countries that demonstrate
relatively better environmental performances as measured by the EPI, but higher air pollution
as measured by CO2 emissions. This adds to our previous discussion about economic complexity
at the country level, as it allows us to understand which sets of products are leading to a better
overall environmental performance and less air pollution, based on their embodied sophistication.
6 Conclusions
Our analysis illustrates that the environmental performance of a country is highly correlated
with the mix of products that it produces and exports. In a panel data setting, we have verified
that there is a robust positive (resp. negative) relationship between environmental performance
27/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
(resp. air quality) and product sophistication. Moreover, the effect of economic complexity on
environmental performance has been verified with fixed-effects instrumental variables estimation
techniques. Thus, the evidence presented in the paper suggests that the sophistication of a
country’s productive structure predicts its environmental performance.
More specifically, countries that produce more complex products are associated with improved
environmental performance but also with inferior air quality (higher exposure to PM2.5 and
CO2 emissions). Explicitly, it seems that when an economy accelerates from an agricultural
productive structure to a more sophisticated one with industrial and technological sectors, the
effect on overall environmental performance is positive but the particular effect on air quality is
negative.
We build a Product Environmental Performance Index and a Product Air Pollution Index that
associate exported products with the average level of countries’ environmental performance as
measured by the EPI and air pollution as measured by CO2 emissions, respectively.
With these indexes, we show how the development of complex products is associated with changes
in the environment. Hence, the two indexes could be a valuable tool for evaluating smart special-
ization strategies and/or sectoral reallocation policies towards activities/sectors that are associ-
ated with better environmental performance and lower air pollution. According to WHO [125],
exposure to air pollution has been found to have both direct and indirect detrimental effects on
human health. The direct effects include respiratory irritation, chronic respiratory symptoms,
heart diseases, lung cancer, premature mortality and reduced life expectancy [37, 70, 94]. The
indirect mechanisms relate to pulmonary oxidative stress and inflammatory responses. Hence,
empirical evidence about the effect of economic complexity on air quality should be highly infor-
mative for policy makers.
In sum, this study identifies economic complexity as an explanatory variable of the observed
differences in environmental footprints across countries. However, we do not attempt to address
28/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
the question of why some countries have a higher level of economic complexity than others.
Trying to answer this question is beyond the scope of this paper, but an interesting way forward.
References
[1] Abdon, A.; Felipe, J. (2011). The product space: What does it say about the opportunities
for growth and structural transformation of Sub-Saharan Africa? Levy Economics Institte
Working Paper (670).
[2] Acaravci, A.; Ozturk, I. (2010). On the relationship between energy consumption, CO2
emissions and economic growth in Europe. Energy 35(12), 5412–5420.
[3] Agras, J.; Chapman, D. (1999). A dynamic approach to the Environmental Kuznets Curve
hypothesis. Ecological Economics 28(2), 267–277.
[4] Aiken, L. S.; West, S. G.; Reno, R. R. (1991). Multiple regression: Testing and interpreting
interactions. Sage.
[5] Akbostancı, E.; Türüt-Aşık, S.; Tunç, G. İ. (2009). The relationship between income and
environment in Turkey: is there an environmental Kuznets curve? Energy policy 37(3),
861–867.
[6] Al-Mulali, U.; Ozturk, I. (2015). The effect of energy consumption, urbanization, trade
openness, industrial output, and the political stability on the environmental degradation
in the MENA (Middle East and North African) region. Energy 84, 382–389.
[7] Albeaik, S.; Kaltenberg, M.; Alsaleh, M.; Hidalgo, C. A. (2017). 729 new measures of
economic complexity (Addendum to Improving the Economic Complexity Index). arXiv
preprint arXiv:1708.04107 .
29/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[8] Albeaik, S.; Kaltenberg, M.; Alsaleh, M.; Hidalgo, C. A. (2017). Measuring the Knowledge
Intensity of Economies with an Improved Measure of Economic Complexity. arXiv preprint
arXiv:1707.05826 .
[9] Ang, J. B. (2007). CO2 emissions, energy consumption, and output in France. Energy
Policy 35(10), 4772–4778.
[10] Ang, J. B. (2008). Economic development, pollutant emissions and energy consumption in
Malaysia. Journal of Policy Modeling 30(2), 271–278.
[11] Antoci, A.; Russu, P.; Sordi, S.; Ticci, E. (2014). Industrialization and environmental
externalities in a Solow-type model. Journal of Economic Dynamics and Control 47, 211–
224.
[12] Antweiler, W.; Copeland, B. R.; Taylor, M. S. (2001). Is free trade good for the environ-
ment? American Economic Review 91(4), 877–908.
[13] Asane-Otoo, E. (2015). Carbon footprint and emission determinants in Africa. Energy 82,
426–435.
[14] Balassa, B. (1965). Trade liberalisation and "revealed" comparative advantage. The Manch-
ester School 33(2), 99–123.
[15] Bank, T. W. (2000). World Development Indicators 2000. Oxford University Press, USA.
[16] Bimonte, S. (2002). Information access, income distribution, and the Environmental
Kuznets Curve. Ecological economics 41(1), 145–156.
[17] Brooks, N.; Sethi, R. (1997). The distribution of pollution: community characteristics
and exposure to air toxics. Journal of environmental economics and management 32(2),
233–250.
30/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[18] Bustos, S.; Gomez, C.; Hausmann, R.; Hidalgo, C. A. (2012). The dynamics of nestedness
predicts the evolution of industrial ecosystems. PloS one 7(11), e49393.
[19] Caldarelli, G.; Cristelli, M.; Gabrielli, A.; Pietronero, L.; Scala, A.; Tacchella, A. (2012).
A network analysis of countries’ export flows: firm grounds for the building blocks of the
economy. PloS one 7(10), e47278.
[20] Carpenter, S. R.; Caraco, N. F.; Correll, D. L.; Howarth, R. W.; Sharpley, A. N.; Smith,
V. H. (1998). Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological
applications 8(3), 559–568.
[21] Chay, K. Y.; Greenstone, M. (2005). Does air quality matter? Evidence from the housing
market. Journal of political Economy 113(2), 376–424.
[22] Chenery, H. B.; Taylor, L. (1968). Development patterns: among countries and over time.
The Review of Economics and Statistics , 391–416.
[23] Cherniwchan, J. (2012). Economic growth, industrialization, and the environment. Re-
source and Energy Economics 34(4), 442–467.
[24] Clark, C. (1967). The conditions of economic progress. The conditions of economic progress.
.
[25] Cole, M. A. (2007). Corruption, income and the environment: an empirical analysis. Eco-
logical Economics 62(3-4), 637–647.
[26] Cole, M. A.; Elliott, R. J. (2003). Determining the trade–environment composition ef-
fect: the role of capital, labor and environmental regulations. Journal of Environmental
Economics and Management 46(3), 363–383.
[27] Coondoo, D.; Dinda, S. (2002). Causality between income and emission: a country group-
specific econometric analysis. Ecological Economics 40(3), 351–367.
31/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[28] Cristelli, M.; Gabrielli, A.; Tacchella, A.; Caldarelli, G.; Pietronero, L. (2013). Measuring
the intangibles: A metrics for the economic complexity of countries and products. PloS
one 8(8), e70726.
[29] Cristelli, M.; Tacchella, A.; Pietronero, L. (2015). The heterogeneous dynamics of economic
complexity. PloS one 10(2), e0117174.
[30] Damania, R.; Fredriksson, P. G.; List, J. A. (2003). Trade liberalization, corruption, and
environmental policy formation: theory and evidence. Journal of environmental economics
and management 46(3), 490–512.
[31] Dietrich, A. (2012). Does growth cause structural change, or is it the other way around?
A dynamic panel data analysis for seven OECD countries. Empirical economics 43(3),
915–944.
[32] Dinda, S. (2004). Environmental Kuznets curve hypothesis: a survey. Ecological economics
49(4), 431–455.
[33] Dinda, S.; Coondoo, D. (2006). Income and emission: a panel data-based cointegration
analysis. Ecological Economics 57(2), 167–181.
[34] Dominici, F.; Greenstone, M.; Sunstein, C. R. (2014). Particulate matter matters. Science
344(6181), 257–259.
[35] Elgin, C.; Oztunali, O.; et al. (2012). Shadow economies around the world: model based
estimates. Bogazici University Department of Economics Working Papers 5(2012), 1–48.
[36] European Commission (1999). The EU’s Eco-Industry’s Export Potential. Final report to
DGXI of the European Commission .
32/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[37] European Environment Agency (2015). Air quality in Europe-2015 re-
port. Tech. Rep. 5. URL http://www.actu-environnement.com/media/pdf/
news-25756-rapport-air-aee.pdf.
[38] Farhani, S.; Mrizak, S.; Chaibi, A.; Rault, C. (2014). The environmental Kuznets curve
and sustainability: A panel data analysis. Energy Policy 71, 189–198.
[39] Feenstra, R. C.; Lipsey, R. E.; Deng, H.; Ma, A. C.; Mo, H. (2005). World trade flows:
1962-2000. Tech. rep., National Bureau of Economic Research.
[40] Felipe, J. (2012). Inclusive Growth, Full Employment, and Structural Change: Implications
and Policies for Developing Asia. Anthem Press.
[41] Foray, D. (2014). Smart specialisation: Opportunities and challenges for regional innovation
policy. Routledge.
[42] Foray, D.; David, P. A.; Hall, B. (2009). Smart specialisation–the concept. Knowledge
economists policy brief 9(85), 100.
[43] Foray, D.; Goenaga, X.; et al. (2013). The goals of smart specialisation. S3 policy brief
series 1, 2013.
[44] Fourastié, J. (1963). Le grand espoir du XXe siècle. Revue Française de Sociologie 4(2),
224.
[45] Frankel, J. A.; Rose, A. K. (2005). Is trade good or bad for the environment? Sorting out
the causality. Review of economics and statistics 87(1), 85–91.
[46] Galeotti, M.; Lanza, A.; Pauli, F. (2006). Reassessing the environmental Kuznets curve
for CO2 emissions: A robustness exercise. Ecological economics 57(1), 152–163.
33/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[47] Galeotti, M.; Manera, M.; Lanza, A. (2009). On the robustness of robustness checks of the
environmental Kuznets curve hypothesis. Environmental and Resource Economics 42(4),
551.
[48] Goel, R. K.; Herrala, R.; Mazhar, U. (2013). Institutional quality and environmental
pollution: MENA countries versus the rest of the world. Economic Systems 37(4), 508–
521.
[49] Grossman, G. M.; Krueger, A. B. (1995). Economic growth and the environment. The
quarterly journal of economics 110(2), 353–377.
[50] Hafner, O. (1998). The role of corruption in the misappropriation of tropical forest resources
and in tropical forest destruction. Transparency International Working Paper .
[51] Halicioglu, F. (2009). An econometric study of CO2 emissions, energy consumption, income
and foreign trade in Turkey. Energy Policy 37(3), 1156–1164.
[52] Hall, R. E.; Jones, C. I. (1999). Why do some countries produce so much more output per
worker than others? The quarterly journal of economics 114(1), 83–116.
[53] Hamilton, C.; Turton, H. (2002). Determinants of emissions growth in OECD countries.
Energy Policy 30(1), 63–71.
[54] Hartmann, D.; Guevara, M. R.; Jara-Figueroa, C.; Aristarán, M.; Hidalgo, C. A. (2017).
Linking Economic Complexity, Institutions, and Income Inequality. World Development
93, 75–93.
[55] Hausmann, R.; Hidalgo, C. A. (2011). The network structure of economic output. Journal
of Economic Growth 16(4), 309–342.
[56] Hausmann, R.; Hidalgo, C. A.; Bustos, S.; Coscia, M.; Simoes, A.; Yildirim, M. A. (2014).
The atlas of economic complexity: Mapping paths to prosperity. Mit Press.
34/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[57] Hausmann, R.; Hwang, J.; Rodrik, D. (2007). What you export matters. Journal of
economic growth 12(1), 1–25.
[58] Hidalgo, C. A.; Hausmann, R. (2009). The building blocks of economic complexity. pro-
ceedings of the national academy of sciences 106(26), 10570–10575.
[59] Hidalgo, C. A.; Klinger, B.; Barabási, A.-L.; Hausmann, R. (2007). The product space
conditions the development of nations. Science 317(5837), 482–487.
[60] Hilty, L. M.; Arnfalk, P.; Erdmann, L.; Goodman, J.; Lehmann, M.; Wäger, P. A.
(2006). The relevance of information and communication technologies for environmen-
tal sustainability–a prospective simulation study. Environmental Modelling & Software
21(11), 1618–1629.
[61] Holdren, J. P. (1991). Population and the energy problem. Population and environment
12(3), 231–255.
[62] Holtz-Eakin, D.; Selden, T. M. (1995). Stoking the fires? CO2 emissions and economic
growth. Journal of public economics 57(1), 85–101.
[63] Hossain, M. S. (2011). Panel estimation for CO2 emissions, energy consumption, economic
growth, trade openness and urbanization of newly industrialized countries. Energy Policy
39(11), 6991–6999.
[64] Hsu, A.; Zomer, A. (2014). Environmental performance index. Wiley StatsRef: Statistics
Reference Online , 1–5.
[65] International Trade Centre (2001). The Environmental Services Business: Big and Growing.
International Trade Forum 2, 6–9.
[66] Jalil, A.; Mahmud, S. F. (2009). Environment Kuznets curve for CO2 emissions: a cointe-
gration analysis for China. Energy policy 37(12), 5167–5172.
35/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[67] Jayanthakumaran, K.; Verma, R.; Liu, Y. (2012). CO2 emissions, energy consumption,
trade and income: a comparative analysis of China and India. Energy Policy 42, 450–460.
[68] Jorgenson, A. K.; Clark, B. (2010). Assessing the temporal stability of the popula-
tion/environment relationship in comparative perspective: a cross-national panel study
of carbon dioxide emissions, 1960–2005. Population and Environment 32(1), 27–41.
[69] Kaldor, N.; et al. (1967). Strategic factors in economic development. New York State School
of Industrial and Labor Relations, Cornell University.
[70] Kampa, M.; Castanas, E. (2008). Human health effects of air pollution. Environmental
pollution 151(2), 362–367.
[71] Kasman, A.; Duman, Y. S. (2015). CO2 emissions, economic growth, energy consumption,
trade and urbanization in new EU member and candidate countries: a panel data analysis.
Economic Modelling 44, 97–103.
[72] Kaufmann, D.; Kraay, A.; Zoido-Lobaton, P. (1999). Governance Matters. World Bank
Policy Research Department. Tech. rep., Working Paper.
[73] Kennett, M.; Steenblik, R. (2005). Environmental goods and services: A synthesis of country
studies, vol. 3. OECD Publishing.
[74] Kuznets, S.; Murphy, J. T. (1966). Modern economic growth: Rate, structure, and spread,
vol. 2. Yale University Press New Haven.
[75] Lall, S. (1992). Technological capabilities and industrialization. World development 20(2),
165–186.
[76] Lapatinas, A.; Litina, A.; Sartzetakis, E. S. (2018). Environmental projects in the presence
of corruption. International Tax and Public Finance , 1–42.
36/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[77] Lee, C.-C.; Lee, J.-D. (2009). Income and CO2 emissions: evidence from panel unit root
and cointegration tests. Energy policy 37(2), 413–423.
[78] Lewis, W. A. (2013). Theory of economic growth, vol. 7. Routledge.
[79] Li, H.; Mu, H.; Zhang, M.; Gui, S. (2012). Analysis of regional difference on impact factors
of China’s energy–Related CO2 emissions. Energy 39(1), 319–326.
[80] Li, K.; Lin, B. (2015). Impacts of urbanization and industrialization on energy consump-
tion/CO2 emissions: does the level of development matter? Renewable and Sustainable
Energy Reviews 52, 1107–1122.
[81] Liddle, B. (2004). Demographic dynamics and per capita environmental impact: Using
panel regressions and household decompositions to examine population and transport.
Population and Environment 26(1), 23–39.
[82] Liddle, B.; Lung, S. (2010). Age-structure, urbanization, and climate change in devel-
oped countries: revisiting STIRPAT for disaggregated population and consumption-related
environmental impacts. Population and Environment 31(5), 317–343.
[83] Lin, C.-Y. C.; Liscow, Z. D. (2012). Endogeneity in the environmental Kuznets curve:
an instrumental variables approach. American Journal of Agricultural Economics 95(2),
268–274.
[84] Lin, S.; Zhao, D.; Marinova, D. (2009). Analysis of the environmental impact of China
based on STIRPAT model. Environmental Impact Assessment Review 29(6), 341–347.
[85] Lippe, M. (1999). Corruption and environment at the local level. Transparency Interna-
tional Working Paper. Retrieved March 31, 2014.
37/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[86] Lisciandra, M.; Migliardo, C. (2017). An Empirical Study of the Impact of Corruption
on Environmental Performance: Evidence from Panel Data. Environmental and Resource
Economics 68(2), 297–318.
[87] Lopez, R.; Mitra, S. (2000). Corruption, pollution, and the Kuznets environment curve.
Journal of Environmental Economics and Management 40(2), 137–150.
[88] Madlener, R.; Sunak, Y. (2011). Impacts of urbanization on urban structures and energy
demand: What can we learn for urban energy planning and urbanization management?
Sustainable Cities and Society 1(1), 45–53.
[89] Marsiglio, S.; Ansuategi, A.; Gallastegui, M. C. (2016). The environmental Kuznets curve
and the structural change hypothesis. Environmental and resource economics 63(2), 265–
288.
[90] Martínez-Zarzoso, I.; Bengochea-Morancho, A.; Morales-Lage, R. (2007). The impact of
population on CO 2 emissions: evidence from European countries. Environmental and
Resource Economics 38(4), 497–512.
[91] Martínez-Zarzoso, I.; Maruotti, A. (2011). The impact of urbanization on CO2 emissions:
evidence from developing countries. Ecological Economics 70(7), 1344–1353.
[92] Mauro, P. (1995). Corruption and growth. The quarterly journal of economics 110(3),
681–712.
[93] Nejat, P.; Jomehzadeh, F.; Taheri, M. M.; Gohari, M.; Majid, M. Z. A. (2015). A global
review of energy consumption, CO2 emissions and policy in the residential sector (with an
overview of the top ten CO2 emitting countries). Renewable and sustainable energy reviews
43, 843–862.
[94] OECD (2016). The Economic Consequences of Outdoor Air Pollution. Tech. rep.
38/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[95] Ollo-López, A.; Aramendía-Muneta, M. E. (2012). ICT impact on competitiveness, inno-
vation and environment. Telematics and Informatics 29(2), 204–210.
[96] Ozcan, B. (2013). The nexus between carbon emissions, energy consumption and economic
growth in Middle East countries: a panel data analysis. Energy Policy 62, 1138–1147.
[97] Panayotou, T. (2016). Economic growth and the environment. The environment in an-
thropology , 140–148.
[98] Pellegrini, L.; Gerlagh, R. (2006). Corruption, democracy, and environmental policy: an
empirical contribution to the debate. The Journal of Environment & Development 15(3),
332–354.
[99] Ranis, G.; Fei, J. C. (1961). A theory of economic development. The american economic
review , 533–565.
[100] Raudenbush, S. W.; Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods, vol. 1. Sage.
[101] Rodrik, D. (2006). What’s so special about China’s exports? China & World Economy
14(5), 1–19.
[102] Romm, J. (2002). The internet and the new energy economy. Resources, conservation and
recycling 36(3), 197–210.
[103] Samet, J. M.; Dominici, F.; Curriero, F. C.; Coursac, I.; Zeger, S. L. (2000). Fine particulate
air pollution and mortality in 20 US cities, 1987–1994. New England journal of medicine
343(24), 1742–1749.
[104] Sari, R.; Soytas, U. (2009). Are global warming and economic growth compatible? Evidence
from five OPEC countries? Applied Energy 86(10), 1887–1893.
39/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[105] Savci, S. (2012). An agricultural pollutant: chemical fertilizer. International Journal of
Environmental Science and Development 3(1), 73.
[106] Saviotti, P. P.; Frenken, K. (2008). Export variety and the economic performance of coun-
tries. Journal of Evolutionary Economics 18(2), 201–218.
[107] Shafiei, S.; Salim, R. A. (2014). Non-renewable and renewable energy consumption and
CO2 emissions in OECD countries: A comparative analysis. Energy Policy 66, 547–556.
[108] Shafik, N. (1994). Economic development and environmental quality: an econometric anal-
ysis. Oxford economic papers , 757–773.
[109] Sharma, S. S. (2011). Determinants of carbon dioxide emissions: empirical evidence from
69 countries. Applied Energy 88(1), 376–382.
[110] Sherbinin, A. d.; Carr, D.; Cassels, S.; Jiang, L. (2007). Population and environment.
Annu. Rev. Environ. Resour. 32, 345–373.
[111] Sinclair-Desgagné, B.; et al. (2008). The environmental goods and services industry. In-
ternational Review of Environmental and Resource Economics 2(1), 69–99.
[112] Soytas, U.; Sari, R. (2009). Energy consumption, economic growth, and carbon emissions:
challenges faced by an EU candidate member. Ecological economics 68(6), 1667–1675.
[113] Soytas, U.; Sari, R.; Ewing, B. T. (2007). Energy consumption, income, and carbon emis-
sions in the United States. Ecological Economics 62(3-4), 482–489.
[114] Stern, D. I. (2002). Explaining changes in global sulfur emissions: an econometric decom-
position approach. Ecological Economics 42(1-2), 201–220.
[115] Stern, D. I. (2004). The rise and fall of the environmental Kuznets curve. World develop-
ment 32(8), 1419–1439.
40/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[116] Stock, J.; Yogo, M. (2005). Testing for Weak Instruments in Linear IV Regression, New
York: Cambridge University Press. pp. 80–108. URL http://www.economics.harvard.
edu/faculty/stock/files/TestingWeakInstr_Stock%2BYogo.pdf.
[117] Tacchella, A.; Cristelli, M.; Caldarelli, G.; Gabrielli, A.; Pietronero, L. (2012). A new
metrics for countries’ fitness and products’ complexity. Scientific reports 2, 723.
[118] Tacchella, A.; Cristelli, M.; Caldarelli, G.; Gabrielli, A.; Pietronero, L. (2013). Economic
complexity: conceptual grounding of a new metrics for global competitiveness. Journal of
Economic Dynamics and Control 37(8), 1683–1691.
[119] Torras, M.; Boyce, J. K. (1998). Income, inequality, and pollution: a reassessment of the
environmental Kuznets curve. Ecological economics 25(2), 147–160.
[120] Tsurumi, T.; Managi, S. (2010). Decomposition of the environmental Kuznets curve: scale,
technique, and composition effects. Environmental Economics and Policy Studies 11(1-4),
19–36.
[121] Wang, C.; Chen, J.; Zou, J. (2005). Decomposition of energy-related CO2 emission in
China: 1957–2000. Energy 30(1), 73–83.
[122] Wang, S.; Fang, C.; Guan, X.; Pang, B.; Ma, H. (2014). Urbanisation, energy consumption,
and carbon dioxide emissions in China: a panel data analysis of China’s provinces. Applied
Energy 136, 738–749.
[123] Wang, S.; Zhou, D.; Zhou, P.; Wang, Q. (2011). CO2 emissions, energy consumption and
economic growth in China: A panel data analysis. Energy Policy 39(9), 4870–4875.
[124] Welsch, H. (2004). Corruption, growth, and the environment: a cross-country analysis.
Environment and Development Economics 9(5), 663–693.
41/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
[125] WHO (2013). Review of evidence on health aspects of air pollution: REVIHAAP project:
final technical report.
[126] Xu, B.; Lin, B. (2015). How industrialization and urbanization process impacts on CO2
emissions in China: evidence from nonparametric additive regression models. Energy Eco-
nomics 48, 188–202.
[127] York, R.; Rosa, E. A.; Dietz, T. (2003). STIRPAT, IPAT and ImPACT: analytic tools
for unpacking the driving forces of environmental impacts. Ecological economics 46(3),
351–365.
[128] Zhang, C.; Lin, Y. (2012). Panel estimation for urbanization, energy consumption and
CO2 emissions: A regional analysis in China. Energy Policy 49, 488–498.
[129] Zhang, X.-P.; Cheng, X.-M. (2009). Energy consumption, carbon emissions, and economic
growth in China. Ecological Economics 68(10), 2706–2712.
[130] Zhou, W.; Zhu, B.; Chen, D.; Griffy-Brown, C.; Ma, Y.; Fei, W. (2012). Energy consumption
patterns in the process of China’s urbanization. Population and Environment 33(2-3),
202–220.
42/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
A Economic complexity indexes: methods
A.1 Economic complexity index (ECI)
To calculate the measures of economic complexity used in this work, we rely on the methodology
described in Hausmann et al. [56]. In short, let us assume that we have trade information for l
number of countries and k products. With this information, we can fill an (l× k) exports matrix
E, so that matrix element Eij is equal to the monetary value country i gains by exporting product
j. Of course, if country i does not export product j, then Eij = 0. From this matrix, it is easy
to calculate the ratio between the share of a given product in a country’s exports and the share
of this product in the total global exports. This ratio is called Revealed Comparative Advantage
(RCA) [14], and is given by
RCAcp =Xcp/
∑p′ Xcp′
∑c′ Xc′p/
∑c′p′ Xc′p′
, (5)
where Xcp is the total value of product p exports by country c. As discussed previously in Cal-
darelli et al. [19], Hartmann et al. [54], Hidalgo and Hausmann [58], a country has a comparative
advantage in a product (in other words, is a competitive exporter of a product) when RCAcp ≥ 1.
Using this threshold value, we obtain the (l × k) matrix M, with matrix elements Mcp = 1 if
country c has an RCA for product p, and zero otherwise. This matrix can be viewed as the
incidence matrix of a bipartite network linking countries to products.
From this matrix, Hidalgo and Hausmann [58] introduced the ECI as a measure of the production
characteristics of different countries. To obtain the ECI, we calculate the (l × l) square matrix
M̃. In short, matrix M̃ provides information about links connecting two countries c and c′, based
on the number of products they both export. The matrix elements M̃cc′ are computed as
M̃cc′ =1
kc,0
∑
p
McpMc′p
kp,0, (6)
43/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
where kc,0 =∑
pMcp measures the diversification of country c in terms of the number of different
products it exports, and kp,0 =∑
cMcp measures the number of countries that export a certain
product p. If K is the eigenvector of M̃ associated with the second largest eigenvalue, then
according to Hausmann et al. [56], the ECI is calculated as
ECI =K− 〈K〉
std(K). (7)
In a similar manner, if instead of countries we place the spotlight on individual products, we can
calculate the Product Complexity Index (PCI). In this case, the (k × k) matrix M̃ will provide
information about links connecting two products p and p′, based on the number of countries that
export them both. Therefore, the matrix elements M̃pp′ are computed as
M̃pp′ =1
kp,0
∑
c
McpMcp′
kc,0, (8)
and if Q is the eigenvector of M̃ associated with the second largest eigenvalue,
PCI =Q− 〈Q〉
std(Q). (9)
A.2 Improved measure of economic complexity (ECI+)
To calculate the improved measure of economic complexity (ECI+) used in this work, we rely
on the methodology described in Albeaik et al. [8]. In short, let us assume that we have trade
information for l number of countries and k products. We can calculate the total exports of a
country corrected by how difficult it is to export each product using
X1c =
∑
p
Xcp∑c
Xcp
X0c
, (10)
44/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
where X0c =
∑pXcp is the total exports of country c and 1/
∑c
Xcp
X0c
measures how difficult it is
for country c to export product p.
We then take this corrected value of total exports (eq. 10) to calculate the second order correction:
X2c =
∑
p
Xcp∑c
Xcp
X1c
, (11)
where X2c represents the share that a product represents of the average country.
Iterating this to the limit:
XNc =
∑
p
Xcp∑c
Xcp
XN−1c
, (12)
and normalizing Xc at each iteration step by its geometric mean:
XNc =
XNc
(∏
c′ XNc′)
1[C]
(13)
where [C] is the number of countries in the sample, we estimate the ECI+ as the total exports
of a country corrected by how difficult it is to export each product, minus the average share that
the country represents in the export of a product (which accounts for the size of a country’s
export economy):
ECI+c = log(X∞
c )− log(∑
p
Xcp
Xp
). (14)
Placing the spotlight on products instead of countries, PCI+ is defined as the following iterative
map:
XNp =
∑
c
Xcp∑p
Xcp
XN−1p
(15)
with the initial condition X0p =
∑c
Xcp
X0c
being the average share of product p in country c.
45/46
A. Lapatinas, A. Garas, E. Boleti and A. Kyriakou:economic complexity and environmental performance
Again, normalizing Xp at each step by its geometric mean:
XNp =
XNp
(∏
p′ XNp′)
1[P]
(16)
where [P ] is the number of products in the sample, we define the product complexity index,
corrected by how difficult it is to export each product, as
PCI+p = log(Xp)− log(X∞
p ) (17)
where Xp is product p’s total world trade.
To summarize, the ECI+ and the PCI+ denote a measure of the total exports of a country,
corrected by how difficult it is to export each product, and a measure of the total trade in a
product, corrected by how easy it is to export that product, respectively [8].
46/46